Written by: Techub News Compilation
Recently, NVIDIA founder and CEO Jensen Huang appeared at a Cisco event after a whirlwind two-week trip to Asia, engaging in an in-depth and lively conversation with Cisco executives. During this exchange, Huang not only systematically explained the core concept of the "AI Factory" as the next-generation smart infrastructure but also offered highly practical advice, from a unique perspective, on how businesses can survive, evolve, and even lead in this AI wave. His views transcended mere technical discussions, touching on fundamental shifts in corporate strategy, innovation culture, and mindset.
AI Factory: Redefining the Infrastructure of Computing
Huang unequivocally pointed out that we are experiencing “the first complete paradigm shift in computing in 60 years.” In the past “explicit programming” era, programmers told computers precisely what to do step by step using languages like Fortran and C++. Today, we are moving towards an “implicit programming” era: you just need to tell the computer your intention, and it can explore how to solve the problem on its own. This shift is not only happening at the processing layer but is sweeping the entire computing stack—storage, networking, security, everything is being reconstructed.
The “AI Factory” is the concentrated embodiment of this new paradigm. It is no longer just a traditional data center or cloud computing platform but an infrastructure capable of continuously producing, refining, and deploying “intelligence.” Huang emphasized that current AI (like early chatbots) may be interesting but is not yet “useful.” True intelligence lies in problem-solving, which requires several key capabilities: knowing what you do not know, reasoning, breaking down unknown issues into solvable subproblems, planning, using tools, conducting research. This is the direction currently being explored in the “AI agent” field.
He further explained that past “retrieval-based” software (like mobile applications) was pre-recorded and static. In the future, software will be “generative,” with each interaction, each prompt, and each context producing unique outputs, just like this conversation between him and the host, where every sentence is appearing in a new combination for the first time. This requires the underlying infrastructure to have unprecedented dynamic computing and data processing capabilities.
Corporate AI Transformation: From "Blooming Flowers" to "Targeted Actions"
When asked how companies should take the first step in AI transformation, Huang gave an intuitive yet profound suggestion: Don’t get caught up in ROI right from the start. He believes that it is difficult to accurately calculate ROI for early deployment of any new technology using spreadsheets. Instead, corporate leaders should ask themselves: What is the essence of the company? What are the core tasks that have the most significant impact?
He shared NVIDIA’s internal practice: “We let a ‘hundred flowers bloom.’” There are thousands of AI projects happening simultaneously within the company, which may seem “out of control,” but this is the norm of innovation. “If you always want to control everything, first you need healing, and secondly, control is an illusion.” He suggested that companies create a safe environment that encourages employees to boldly experiment with various AI tools—whether it’s Anthropic, Codex, Gemini, or any other model. When a team expresses a desire to try an AI technology, his first reaction is “Okay, then I’ll ask why,” instead of “Prove it to me first.” This attitude mirrors how he educates his children: encouraging exploration rather than preset limitations.
Of course, after the “hundred flowers bloom,” it is necessary to “prune the garden.” Companies need to exercise judgment at some point, start integrating resources, and focus on the most promising platforms or methods. Huang revealed that he has not yet begun large-scale “pruning,” but he is very clear about what the most important core tasks for the company are. For NVIDIA, this is undoubtedly chip design, software engineering, and systems engineering. Therefore, he ensures that the company invests heavily in experts and resources to leverage AI to completely innovate workflows in these areas and collaborates deeply with tool giants like Synopsys, Cadence, and Siemens to seamlessly integrate NVIDIA’s technology into the design process.
He proposed a key mindset shift: Examine your business through the lens of “infinite computing power”. Imagine if the time taken for a job could be reduced from a year to a day, an hour, or even completed in real-time; if processing massive data (like trillion-scale graph analysis) no longer required chunking but could be processed in one go; if the constraints of “quality” or “speed” no longer existed. If you haven’t approached your company’s toughest problems with this “infinite/zero-cost” logic, you may be doing it wrong. He warned, “You may not think this way, but just imagine your competitors are thinking exactly this way, or a newly formed company is thinking this way, everything changes.”
Technology Stack, Physical AI, and the "Digitalized" Future of Enterprises
Huang likened the AI technology stack to a “five-layer cake”: from the bottom layer of chips and power, to infrastructure (hardware and software), and then to AI models, but the topmost and most important layer is always applications. “For God’s sake, go apply this technology!” he urged, “A company using AI will not fall into crisis. You won’t be replaced by AI; you will be replaced by those who use AI.”
He rebutted the popular notion that “AI will replace software companies,” calling it “the most illogical idea in the world.” He illustrated with a thought experiment: if a general humanoid robot (AG Robotics) needed to screw in a screw, would it reinvent the screwdriver? Of course not; it would use the existing screwdriver. Similarly, a digital world AGI would use existing software tools (like ServiceNow, SAP, Cadence) rather than reinventing the calculator. This is precisely why “tool usage” in the AI field has become a breakthrough development. Many problems in the physical world have clear laws (like F=ma, V=IR), and AI should learn to use tools built for these laws instead of re-deriving them.
This leads to his vision of “physical AI.” The next generation of physical AI needs to understand the physical world and causality. He used “dominoes” as an example: a child can easily understand that knocking over one domino will cause a chain reaction; this concept integrates complex principles like causality, contact, gravity, and mass, which current large language models struggle to grasp. Creating AI capable of understanding such physical concepts will be a tremendous opportunity.
Huang proposed an exciting viewpoint: AI will help every company transform into a “technology-first” company. “I love Disney, but I bet they would rather be Netflix. I love Mercedes-Benz (I came in one), but I’m sure they would rather be Tesla. I love Walmart, but I’m sure they would rather be Amazon.” He cited these three examples to point out that traditional companies aspire to be the “technology-first” players in their fields. Technology should become their superpower, while domain knowledge represents application scenarios. Because technology companies deal with “electrons,” not “atoms,” the scale of electrons is almost limitless, which can transcend the physical limitations of the atomic world.
He encouraged all businesses to embrace this transformation: “At its core, programming (coding) is just typing, and typing has become a commodity. You all possess an incredibly valuable asset—domain expertise. You understand your customers and know where the problems lie. Now, for the first time, you can use your own language to directly tell the computer what you want.” This means that the greatest value of enterprises—intent and understanding—will be directly transformed into productivity like never before.
Private Deployment, AI Integration, and a Call to Action
Regarding whether enterprises should choose public cloud or build their own infrastructure, Huang offered a passionate answer: “Go build a computer!” Even though PC technology is mature and widespread, it is crucial to construct and understand every component personally. “If you are in the transportation industry, don’t just use Uber. Open the hood, change the oil, and understand how it works.” This hands-on experience is vital for understanding this technology that will define the future. He believes the future will be a mixed model of “partially leased, partially owned.”
He particularly emphasized the importance of data sovereignty and privacy, illustrating with a vivid metaphor: “When I go to see a psychologist, I do not want my problems to be posted online.” For a company, the most valuable intellectual property often lies not in the answers but in the questions themselves. “What problems am I thinking about, what is important to me—I don’t want others to know this.” Therefore, AI applications involving core strategic dialogues and uncertainties should be deployed in local, private environments. NVIDIA itself has built local super AI systems because he is not comfortable putting all company dialogues in the cloud.
Finally, Huang overturned a common concept: “The idea that ‘AI should always have a human in the loop’ is completely wrong and inverted. It should be every company must have AI in the loop.” He explained that companies wish to become better, more valuable, and richer in knowledge every day, never going backward or starting from scratch. If AI is integrated into business processes, it will capture the company's collective experience. In the future, every employee will have AI assistance, and these AIs will become the core intellectual property that the company continues to accumulate and evolve. This is what the future company will look like.
At the end of the conversation, with a mix of humor and sincerity, Huang gave the most direct call to action: “I think you all should immediately call Chuck (Cisco executive). Or you can call Jensen (myself).” This dialogue, filled with profound insights, strategic foresight, and a touch of wine-tasting cheer, concluded amid the audience’s laughter and applause, leaving endless thoughts on the survival and development path for businesses in the AI era.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
